Erin Purnell Computer Security – Assignment 2 The purpose of this paper is to discuss the different biometric technologies used in security, access control, and identification. Biometrics is a way of identifying an individual from others based on human traits. There are physiological and behavioral characteristics which include, but are not limited to, fingerprinting, palm printing, DNA, iris and retina scans, scent, face recognition, voice, and gait. We will discuss a few of these and what makes each of them better or worse than the others based on a set of standards we will develop. Things to consider would be speed, accuracy, affordability, possibility of errors, invasiveness, etc. More common methods of access control and security are identification cards, passwords, pin numbers, passports, and the ability to answer certain specific questions. Currently we are seeing this technology become must more common in companies’ hiring processes, amusement parks, high security buildings, concerts, and many more public situations. Biometrics is more accurate, reliable, and secure since these tests are unique to each person and difficult to replicate or imitate. However, the use of fingerprinting and DNA analysis poses the issue of privacy and anonymity since this information must all be stored in databases to be able to verify and compare against.1 The governments of countries that utilize this technology will most likely not release all capabilities and information on how the data is being collected and what it could be used for.2 While this type of security has not yet been more widely implemented than identification cards and passwords, it is very likely that companies and technologies begin to adopt these principles and integrate into their policies and products. For example, many smart phones and PCs are beginning to include a fingerprint or retina scanner feature for easy login.3 While these features can be turned off on your personal devices, if required to enter a building or obtain a job, ethical questions will have to start being asked and likewise, answered. A part of the discussion will address these types of issues, but the discussion will mainly attempt to analyze the different types of biometric data and identification and which is most useful for its intended purpose: security. The system we will use for comparing different biometrics and making an informed intelligent decision as to which is “best” to use, will be composed of different talking points and concerns from the business’/implementers’ side and the end-users’ side. Since security is the foremost concern, it is safe to say that accuracy is the most important thing to consider when choosing the best biometric technology 1 "Biometrics security solutions". sourcesecurity.com. Retrieved 22 August 2013. "Biometrics: Overview". Biometrics.cse.msu.edu. 6 September 2007. Jain, A., Hong, L., & Pankanti, S. (2000). "Biometric Identification". Communications of the ACM, 43(2), p. 91-98. DOI 10.1145/328236.328110 Jain, Anil K.; Ross, Arun (2008). "Introduction to Biometrics". In Jain, AK; Flynn; Ross, A. Handbook of Biometrics. Springer. pp. 1–22. ISBN 978-0-387-71040-2. 2 Defense Science Board (DSB) (September 2006). "Chapter 17, Recommendation 45". Unclassified Report of the Defense Science Board Task Force. Washington, D.C. 20301-3140: Office of the Under Secretary of Defense For Acquisition, Technology, and Logistics. p. 84. 3 Get a reference for identifying unique users. Since many aspects of humans’ biology are so similar from person to person, if a test was inaccurate, there would be no use for it and would also allow for fraud. Reproducibility, while a very important aspect, cannot be separated from accuracy. If a test is accurate consistently—and that is what is meant by accuracy—then it must be reproducible. There is a slight difference in this case though; if a test is accurate so long as the data does not change, reproducibility is not a concern. However, if someone has a fingerprint that is altered by a scar, while the test itself may be accurate, it becomes difficult to reproduce in that situation—this will be addressed specifically within the fingerprint/palm/iris/retina scan sections. I would also suggest combining affordable and available together since equipment that is difficult to obtain is more than likely expensive. There is a difference here though too—there may be equipment that is expensive but prevalent or cheap but rare. In either case, if a company or person is set on one type of technology, price may not be much of an issue. Likewise, if a company or person is trying to spend only a certain amount of money, the biometric test itself must not be of much concern. It is when both are taken into consideration that we must analyze the value of the test. For this reason, we’ll keep them as one validation point. Speed is a very necessary point to consider. If it takes someone 20 minutes to log into their computer because the biometric analysis software is that slow, that would likely not be welcomed. If it took a person thirty minutes to validate that they are in fact who they are so that they can enter an amusement park or their work office, this would be a very serious inconvenience and waste of time. Users of these technologies, especially those that are required involuntarily, should not be inconvenienced so severely. We would hope that this fantastic technology would be more convenient than carrying ID cards or remembering specific information since it certainly is more secure. The analysis should be non-invasive for the consumer, non-embarrassing, and not harmful. It should also not contain/store too much information about the biometric that is obtained. For example, a test that resulted in an entire medical diagnosis being available to the system would not be welcomed by many users. A person would most likely not want to have to urinate in public to get into an office building or get pricked in the finger to give a blood sample. These tests must be convenient and also only give out the information necessary to identify them from someone else, not release personal information into the database of the facility—which could be hacked. For these reasons, we will combine non-invasive, non-embarrassing, and non-harmful into one category and consider it last on the list of qualifications since, again, we are most concerned with security for the sake of this argument. There is an article that also lists universality—meaning that every user needs to possess the trait, uniqueness—meaning that the trait itself must be unique enough between individuals to identify one from the other, permanence—meaning that the trait must not change much over time (the fingerprinting scar example given earlier), measurability—referring to the ease of collection (urine example above), performance—speed of the test (mentioned above), acceptability—individuals being willing to participate (mentioned above when talking about invasiveness, etc), and circumvention—referring to ease of replication (security).4 The only one we had not mentioned previously is that the trait itself must be universal, but that is assumed in our accuracy discussion. These factors must all be considered when deciding which technology is best and they will be 4 Jain, A.K.; Bolle, R.; Pankanti, S., eds. (1999). Biometrics: Personal Identification in Networked Society. Kluwer Academic Publications. ISBN 978-0-7923-8345-1. addressed within our discussion of the other factors for each of the biometric tests we will analyze. The guide for decision making will then be as follows, always with security in mind: Accuracy and Reproducibility - Circumvention of trait - Uniqueness of trait - Permanence of trait Affordability and Obtainability of equipment Speed - Performance of test Non-Invasiveness, Non-Embarrassing, Non-Harmful - Measurability of trait - Acceptability of test We will discuss 5 categories of traits that could be considered most common to these types of test. They are all mentioned in the introduction of this discussion but have been categorized a bit. First is finger-printing and palm-printing. We will compare and choose a winner. DNA with contain blood and saliva. A winner will be chosen. Face recognition will be considered independently. Iris and retina scans will be considered and a winner will be chosen. Voice will be considered independently as well. We can then choose an overall winner for “best” biometric test to use when attempting to be secure and perform as best as possible in the above four validation categories. Finger-printing and palm-printing are probably the most common technology used. These two methods have been being used since it was possible to dip the finger or hand into any type of ink or paint and transfer the image onto some kind of paper. Digital fingerprinting has only become common in recent years but it much more accurate with the use of 3D digital finger printing as of 2010. No two finger prints have ever shown to be identical (uniqueness has not been proven though) which means that the accuracy of finger and palm print analysis is very high. The only situation where someone would not be identified as themselves is in the case of a scar or mutilation to the print, in which case their profile would need to be updated or a court decision would have to be “null.” However, we do need to consider that fingerprints are prevalent where a person resides or spends a significant amount of time and are easily picked up with a material as common as tape. Fingerprints can be planted at crime scenes or copied to some sort of material and used for entry into secure systems and buildings. I would think obtaining a fingerprint would be much easier than an entire palm for circumvention. I would suggest that palm prints have a bit more security than finger prints because the lines on the palm are just as unique as the finger so it would follow that more unique lines would mean more secure analysis. Finger and palm printing are non-invasive, non-embarrassing, and only harmful in the event that ink or paint was used and there was some sort of allergy involved. DNA can be obtained from fingerprints in some situations but if the test is strictly printing, exposing too much personal information is irrelevant. The speed of the test itself is rather fast when taking the actual print. The speed of the comparison to preexisting profiles for that particular person would depend on the system’s speed but would be the fastest of all tests since it would just be a comparison of images. The printing itself can be a bit tedious—the person taking the actual print must be careful of sweat, dirt, etc. and friction so that the print comes out as clearly as possible. Also the print can and will appear slightly different each time it is taken depending on the outside conditions which might cause a bit of stress on the machine doing the comparison. This might be the most cautious process of all of the biometric tests since it involves human error. Regardless, the technology is probably the easiest to obtain and fairly non-expensive (as little as $150)5 other than the actual database where the information is held based on size capacity. We will declare the fingerprint to be the winner of the two in this situation based on the sole reason that it is much easier to obtain a clear, accurate image of each finger individually than it is to get the entire palm at once clearly. Otherwise, the two are tied since you can do each finger individually and store to one user file and obtain the same amount of prints per individual. DNA extraction is next. The two examples we’ll look at are blood and saliva. Accuracy and reproducibility are extremely high in this case. A person’s DNA does not change over time and is 100% different than all other individuals other than the rare case of identical siblings. There can be machine error in extracting the DNA from the samples themselves, but this process is all but perfect. Machines running DNA analysis generally take a few hours which would be a serious problem if you needed to enter your home or login to your computer. This wait is only worth it in court cases, profiling, medical testing, etc. because of its accuracy and telling abilities about the subject—which poses another one of our issues. The amount of personal information that can be obtained from DNA is more than most would want to expose. There are plenty of mental diseases and disorders and other embarrassing and incriminating (in some cases) that a person could have and want to keep private that would all be on display through DNA analysis. Yes, only the match is what is being sought after when we speak about security, but all of the information is stored in those databases regardless of the purpose of the database. If a storage facility containing that information would ever be broken into, the amount of personal information extracted would be immense and extremely invasive. Also, the government is fully capable of requesting DNA database information from not just this country, but other countries if they claim it is a “national security concern”—or something similar. Because DNA contains such valuable information, almost all databases are known and can be taken without the individuals being notified. The cost of DNA equipment can range from a few hundred to a few thousand dollars.6 The equipment is very obtainable—a simple Google search returns thousands of results ranging from home use machines to FBI-worthy equipment. The big issue with DNA testing is that it can be extremely invasive in more than a few ways. The unlikelihood of someone wanting to swab the inside of their cheek every time they enter their building goes without saying. While obtaining saliva is not harmful, obtaining blood in every circumstance is. At the very least, the individual would have a small hole on the finger and at most, vials of blood taken. This is not efficient in time or extraction but is extremely efficient in accuracy and security. Unfortunately, it is very easy to plant DNA at crime scenes or to bring DNA to test as someone other than yourself. Blood and saliva are not necessarily easy, but hair, skin, sweat, etc. are very easy to obtain and plant. For the reasons discussed above, all forms of DNA (skin, blood, saliva, etc.) are all equally accurate but have different levels of embarrassment and harmfulness. When speaking directly 5 6 Gokeyless.com Acelabsystems.com about blood and saliva, saliva is the obvious winner as far as harm goes but both are tied in all other aspects. There is a good chance of DNA being compromised from environmental elements and can be easily contaminated if the proper precautions are not taken. Face recognition is sufficiently accurate when the image is taken full frontal or from 20 degrees to each side.7 There are weaknesses though when a more extreme angle has to be used. Doing facial feature analysis with skin texture analysis increases accuracy 20-25%.8 Face recognition can be analyzed using a variety of algorithms including eye placement, skin texture, etc. It also relies heavily on the lighting of the image and things the individual might be wearing or using to cover parts of their face. Facial expressions can also alter the result.9 Also, it has been shown that sometimes just the installation of a camera system linked to a facial recognition machine can be more effective than the analysis itself because criminals will avoid that area for fear of being detected. The machines necessary can range in price from low to high depending on the functionality of the camera to the software for detection that it is difficult to quantify. Speed works in a similarly dependent-on-the-technology way. While the facial recognition isn’t particularly embarrassing itself, usually these are implemented in public areas where multiple people can be profiled at the same time, while sometimes doing embarrassing actions—not thinking they are being watched. Being able to capture multiple people at once is definitely a plus, though. A definite benefit to facial recognition is that it is impossible to plant another person’s face as evidence or bring another person’s face to pass as your own—unless the fraud itself is a crime (something seriously disturbing). With security cameras installed, faces can be obtained without the individuals’ consent which separates this type of biometric from most others that have been discussed and it not necessarily a good thing when speaking of security. If a machine can run facial recognition off of a picture of another person’s face, the security of actually having to be the person you are claiming to be is nullified. As far as personal security and access control go though, facial recognition seems to be a fairly good solution. Next we’ll discuss iris recognition and retinal scans. The two are often confused but are very different. Retinal scans look at the pattern of blood vessels at the back of the eye. Every individual has a unique and permanent retina, even in the case of identical twins. While these blood vessel patterns are mainly permanent other than in the case of diabetes and glaucoma. This makes it the “most precise and reliable biometric, aside from DNA.”10 The error rate for this analysis is as low as 1 in a million. This type of test is extremely fast—the scan takes seconds and the analysis takes only seconds as well. However, there are several minutes that an individual must spend in darkness to allow the pupil to open. It is absolutely not reproducible unless someone were to have the eye of another person—another unspeakable situation. The technology is absolutely non-embarrassing and does not contain much information other than the image itself and the person’s identity relative to prior records. The only possible issues are that a person may have light-sensitivity and the infrared light that is projected into 7 Williams, Mark. "Better Face-Recognition Software". Williams, Mark. "Better Face-Recognition Software". 9 Bonsor, K. "How Facial Recognition Systems Work". 10 Retina and Iris Scans. Encyclopedia of Espionage, Intelligence, and Security. Copyright © 2004 by The Gale Group, Inc. 8 the individual’s eye could be painful and that some perceive this to be invasive because of the close proximity to the camera equipment. The cost is a bit high, usually in the thousands.11 Besides the invasiveness and cost, this is a very accurate and secure technology. Iris recognition is the identification of an individual based on the complex patterns on the persons’ iris (colored part of the eye). The rate of false match is “infinitesimally small.”12 The pattern of the eye is more permanent as than retina which is extremely beneficial, but the uniqueness has not been proven, much like the fingerprint. Much like the retina though, even identical twins do not have identical irises since they are formed during gestation no genetically. This type of test is more common than a retinal scan and the databases housing these images can be searched at a rate of millions of templates per second. This type of test is much less invasive and harmful than the retina scan since the images can be taken up close or a few feet away; there is no necessity to be close to equipment or have a light projected into your eye. It also cannot reveal any personal information just based off of this test. The scanners can be fooled though with a high-quality image where a retina scan cannot. The iris scanners are relatively expensive much like the retina scanners. The short comings of retina scans are that they are invasive and can change over time with disease. The short coming of iris recognition is that the iris can change pattern depending on how dilated the eye is at the time and also, you can fool the camera with a high-quality picture. This will definitely be the most difficult distinction but when it comes to security, circumvention is a serious issue and while potential changes in retina patterns can—but most likely will not—occur, the risk for someone posing as another is high which defeats the purpose of security for access control and identification. The winner here will be retina scans and we will have to live with the potential invasiveness. Lastly, we’ll talk about voice recognition—also referred to as speaker recognition since it is identifying who is talking not what is being said.13 There are two different types: verification and identification; we will not distinguish between the two for the validation points. There are also textdependent and text-independent tests. Both are just as accurate but the text-dependent case can ask for other identifying information. The text-independent test can allow for an invasive recording of the person’s voice without their knowledge. The test is fairly accurate but can be affected by an individual with a cold, background noise, voice changes over time with age, etc. Also, if a high quality recording was played into a speaker, falsification is easily achievable making access control and identity verification unsecure. It is extremely fast and can be performed in less than 30 seconds.14 To be able to differentiate between a digital voice (recording) and a real voice requires a highly trained forensic expert or very expensive software. It harmless, non-embarrassing, and does not reveal personal information. 11 Timeclockeshop.com En.wikipedia.org/wiki/iris_recognition 13 Kinnunen, Tomi; Li, Haizhou (1 January 2010). "An overview of text-independent speaker recognition: From features to supervectors". Speech Communication 52 (1): 12–40. 14 Matt Warman (8 May 2013). "Say goodbye to the pin: voice recognition takes over at Barclays Wealth" 12 The best of each category are as follows: fingerprinting, saliva (DNA), face recognition, retina scan, voice recognition. Now we must compare them. I have compiled a chart below to summarize what has been written above: (x denotes an issue, o denotes efficiency) Accuracy Fingerprinting DNA Face recognition Retina scan Voice recognition O O X O X Affordability & Obtainability Medium High Medium High Medium Speed Invasiveness O X O O O O X O X O Overall security Medium High Medium-low High Medium-low We consider face and voice recognition to be low in accuracy since various changes and angles can affect accuracy. We consider the cost of DNA and retina scans to be an inconvenience. We consider the speed of DNA to also be a serious inconvenience. The invasiveness of DNA and retina scans are an issue, DNA much more than retina. As far as overall security goes, face recognition, voice recognition, and fingerprinting are the most replicable of the 5. Since we are to consider security first and foremost, DNA and retina scans are the two winners. Between these, retina scans win because of the speed which is very important in high-profile cases, access, and identification processes. While we’d like to choose fingerprinting because of its ease, it is difficult to overlook how simple planting or picking up a fingerprint for falsification can be. The retina scan and iris recognition distinction was also a difficult one but because of the threat of falsification, retina must be chosen as far as security and accuracy go. So while retina is our winner, I think a full assessment of some of the most common biometric technologies shows that most are fairly accurate and inexpensive. They can be implemented at business and homes alike.